Effective oil reservoir management requires the ability to accurately estimate what the current and future performance of existing wells should be and to predict the performance of wells in undeveloped areas. This requires a continuing process of geologic and engineering study, which has led to the use of numerical reservoir simulation models. These models are a powerful tool for analyzing the production history of a reservoir and for predicting future performance under a variety of possible operating methods.
In order to create an acceptable model an accurate reservoir description is essential. This is developed through available geologic and engineering data. The reservoir description is then incorporated into a computer simulation model which is used to verify and further refine the reservoir description through matching of pressure and production data. The refined simulation model is used to project future reservoir performance under current operations and to evaluate alternative operating plans.
Reservoir descriptions may be produced in a number of different ways. For example, in one approach core data are combined with the geologic reservoir description to determine the three-dimensional distribution of porosity and permeability throughout all the geologically defined layers of the reservoir, and a three-dimensional grid system incorporating the layers is used in the model. Because of the amount of resolution in the vertical dimension caused by the definition of the geologic layers the areal grid in this procedure has to be kept relatively coarse so as not to cause excessive computer run time. In any event, despite the grid size utilized, the use of a computer program to simulate performance in a large field requires that a great deal of time and effort be spent in gathering, verifying and loading data. It also requires an amount of computer capacity and available computer time which is normally beyond the means of smaller companies and which is sought to be minimized by the larger companies.
In another approach the field is divided into segments each of which represents an area enclosed within no-flow boundaries. Simulations are conducted for each segment independent of the other segments. The response for each segment can then be pieced together to obtain the response for the entire field or similar field segments are collected into groups with like characteristics, with a single prototype being assigned to each group. Aside from questions which may be raised as to the accuracy of the method, this approach requires the development of models for the entire field under study, which again requires great deal of computer capacity and computer time.
In another approach a model is built by comparing logging data with computer-simulated subsurface characteristics within discretely-boundaried beds. This approach, like the others described, requires actual data from the entire field it is desired to model, and so suffers from the same problems of data entry demands and computer expense and availability.
Regardless of the specific manner of building a computer model, in an effort to reduce the amount of computer simulation required it is common to select a certain relatively small number of wells which can be handled by the available computer capacity and to use them as a window area for the rest of the field. Thus a model derived from these selected wells would be used as an indicator of performance for larger areas of the field. Even this approach, however, can be very expensive and not practical for a relatively small company of limited means.
It is apparent that a different type of reservoir simulation which does not have the great computer demands of the prior art methods is needed.